An intelligent power regulation method and system for a distribution box based on artificial intelligence

Abstract

This invention relates to the field of power regulation technology, and more specifically, to an intelligent power regulation method and system for distribution boxes based on artificial intelligence. The method includes: collecting the current and voltage of each branch circuit and the local temperature between the casings of adjacent intelligent circuit breakers; obtaining the effective value of the current and multiple harmonic components based on the current and voltage, and constructing a standardized data frame together with the local temperature; inputting the standardized data frame into a classification model to obtain the load type; dynamically labeling each branch circuit as a flexible load or a rigid load according to the load type and communication handshake results; identifying at least two consecutively installed intelligent circuit breakers on the guide rail as adjacent groups, and setting a first temperature threshold for each adjacent group; and calculating the thermal contribution of each branch circuit within the group in response to the local temperature exceeding the first temperature threshold. This invention effectively prevents the risk of local heat accumulation and thermal coupling-induced false tripping in enclosed spaces, and ensures the user's continuous power supply experience while avoiding direct power outages.

Description

Technical Field

[0001] This invention relates to the field of power regulation technology. More specifically, this invention relates to an intelligent power regulation method and system for distribution boxes based on artificial intelligence. Background Technology

[0002] As the core control hub in the power distribution network at the end of the power system, the distribution box is mainly responsible for safely distributing the main incoming power to each power branch. Inside, multiple miniature circuit breakers are usually densely and side by side installed on the guide rail to realize overload and short circuit protection for each branch, so as to ensure the physical isolation and operational safety of the electrical systems of buildings such as residential buildings, commercial buildings and industrial plants.

[0003] In recent years, with the popularization of electric vehicle charging piles, high-power variable frequency air conditioners and various smart appliances, the electrical load connected to the terminal distribution box has become highly complex and volatile. In the small and enclosed physical space of the distribution box, when multiple high-power devices operate simultaneously or frequently start and stop, a typical high-density complex power consumption scenario is formed. In order to prevent the distribution box from being overloaded, the existing technology generally adopts a single indicator based on the total incoming power of the distribution box for monitoring. With the help of a static priority mapping table, once the total capacity is predicted to exceed the limit, a disconnection command is sent directly to the branch with lower priority to perform physical power outage.

[0004] However, when using the above-mentioned traditional mechanisms to deal with high-density and complex power consumption scenarios, the existing practices completely ignore the local thermal effects in the enclosed space of the distribution box, resulting in serious physical damage and abnormal user experience. Specifically, when adjacent circuit breakers are carrying continuous high current and high-frequency start-stop loads respectively, even if the total power of the entire box is far from reaching the tripping threshold of the main circuit breaker, local areas will still experience severe heat accumulation due to lack of heat dissipation space. This localized continuous high temperature will not only accelerate the electrical aging of the circuit breaker's insulation shell and internal conductive components, but will also cause the thermal magnetic tripping device of the adjacent circuit breaker to malfunction and trip before reaching its own rated overload current through heat conduction. Summary of the Invention

[0005] This invention provides an intelligent power regulation method and system for distribution boxes based on artificial intelligence, aiming to solve the problem in related technologies where local areas of adjacent circuit breakers also suffer from severe heat accumulation due to lack of heat dissipation space, which can cause the thermal magnetic tripping devices of adjacent circuit breakers to malfunction and trip before reaching their rated overload current through heat conduction.

[0006] In a first aspect, the present invention provides an intelligent power regulation method for a distribution box based on artificial intelligence, comprising: S101: collecting the current and voltage of each branch and the local temperature between the casings of adjacent intelligent circuit breakers; obtaining the effective value of the current and multiple harmonic components based on the current and voltage, and forming a standardized data frame together with the local temperature; S102: inputting the standardized data frame into a classification model to obtain the load type; dynamically labeling each branch as a flexible load or a rigid load according to the load type and communication handshake results; S103: determining at least two consecutively installed intelligent circuit breakers on the guide rail as adjacent groups, and setting a first temperature threshold for each adjacent group; in response to the local temperature exceeding the first temperature threshold, calculating the thermal contribution of each branch in the group, wherein the thermal contribution is positively correlated with the square of the effective value of the current, the standard deviation of the effective value of the current within a set window, and the number of start-stop cycles per unit time; S104: selecting the branch with the highest thermal contribution and labeled as a flexible load as the regulation object, and sending a derating command to it; the derating command includes a target limiting current; S105: the terminal device accordingly reduces its own current until it recovers to the reference current before suppression. In application scenarios where the distribution box is a small, enclosed space and multiple types of complex loads are concentrated, the transient electrical characteristics of the branch circuit and the local temperature of the micro-zone of the adjacent circuit breaker are collected simultaneously. By using an artificial intelligence classification model combined with protocol communication handshake, flexible loads with adjustable potential are dynamically identified. Then, when the local thermal effect exceeds the limit, the true thermal contribution is calculated by comprehensively considering multiple dimensions such as current steady state, fluctuation and start-stop frequency. The core heat source that causes local thermal coupling is accurately located and smooth capacity reduction is implemented. This not only eliminates the safety hazard of adjacent circuit breakers tripping due to heat conduction from the root, but also ensures the user's continuous power experience to the greatest extent without cutting off the physical power supply.

[0007] Furthermore, it also includes: setting a second temperature threshold for each adjacent group, the second temperature threshold being higher than the first temperature threshold; in response to the local temperature exceeding the second temperature threshold, entering an aggressive suppression process, the target current reduction of the derating command is twice that of the conventional suppression process. A progressive dual-temperature threshold safety protection system is constructed to address heat accumulation risks of varying severity. When the local micro-area temperature approaches the second high-temperature critical point that may cause irreversible aging of the circuit breaker insulation casing, the system can adaptively trigger an aggressive suppression strategy with doubled derating, thereby rapidly suppressing the malignant heat source with stronger intervention under extreme heating conditions, avoiding permanent hardware damage caused by the slow conventional derating speed.

[0008] Furthermore, restoring the reference current to the level before suppression includes: when the local temperature is lower than the product of a first temperature threshold (a preset multiple) and 0.85 and this condition persists for a cooling period, the recovery process is initiated, gradually increasing the current with a preset recovery step size until the reference current is reached. In the power recovery phase after the local thermal risk is eliminated, an innovative approach is adopted, introducing temperature hysteresis range and forced cooling period parameter constraints based on a specific multiple ratio. The load current is then slowly released with a preset fixed step size, effectively overcoming the system oscillation problem caused by traditional single-point threshold control, which easily leads to repeated load adjustments near the critical temperature, thus ensuring a smooth transition of the distribution network load state.

[0009] Furthermore, it also includes: if the local temperature exceeds the first temperature threshold again during the recovery process, the recovery will be immediately paused and the current value of the previous level will be returned. The dynamic closed-loop monitoring mechanism during the load power recovery process has been further improved. By continuously detecting the temperature backfire phenomenon during the recovery period and imposing the instruction constraint to immediately return to the current of the previous level, the secondary thermal over-limit caused by the physical heat dissipation lag is precisely prevented, ensuring that the distribution box is always within the safe thermal balance envelope.

[0010] Furthermore, the communication handshake process includes: the main control module sending a protocol handshake message to the identified branch end device; receiving a response message containing the correct device identifier and passing CRC check within a 500ms timeout period; determining it as a flexible load and recording its available protocols; otherwise, temporarily downgrading it to a rigid load. Based on the initial model classification, a communication protocol handshake detection step with strict timing and frame format verification requirements is added. This effectively identifies and filters out pseudo-flexible devices that cannot respond due to controller failure or interface damage. By dynamically downgrading them to unadjustable rigid loads, the system avoids blindly issuing useless derating commands to the out-of-control end, significantly improving the determinism of the underlying execution logic and the reliability of the field.

[0011] Furthermore, it includes a fallback protection mechanism: when the local temperature of any adjacent group exceeds the first temperature threshold but there is no flexible load in the group, an alarm is triggered; when any flexible load fails to perform adjustment according to the derating command, causing the local temperature to continue to rise to the second temperature threshold, the mechanical tripping mechanism of the corresponding smart circuit breaker is activated to forcibly disconnect the branch. By constructing a multi-level system fallback protection network, in response to extreme abnormal conditions such as the absence of adjustable objects in adjacent groups or severe loss of control and refusal to operate of flexible loads, alarms and forced physical isolation measures based on the mechanical tripping mechanism of smart circuit breakers are decisively activated. This ensures that even when the flexible adjustment measures at the software level completely fail, the last line of defense against distribution box fires and electrical collapse can still be maintained.

[0012] Furthermore, the classification model consists of three one-dimensional convolutional layers, two levels of max pooling layers, one flattening layer, and two fully connected layers. This lightweight neural network architecture, employing a specific parameter ratio of three one-dimensional convolutional layers combined with two levels of max pooling, significantly reduces processor computational power requirements and memory storage overhead while still efficiently capturing deep load attributes in transient current timing and harmonic characteristics.

[0013] Furthermore, the target limiting current needs to satisfy the following relationship: In the formula, The target current limit for the selected branch, i.e., the derating command. The effective value of the current in the selected branch; The minimum operating current limit for the selected branch; The derating step size is defined by a formula that clearly defines the safety boundary between the target limiting current and the minimum operating limit of the equipment. While forcibly reducing the heat generation power of the branch, it strictly ensures that the flexible terminal equipment can maintain the most basic standby or minimum maintenance operation state, preventing unexpected equipment shutdown or damage to core components due to excessive input current.

[0014] Furthermore, the method for obtaining the reference current includes: before issuing the adjustment command, the effective value of the instantaneous current before the suppression action is triggered needs to be stored in a cache as the reference current. Using the actual instantaneous effective current at the moment the suppression action is triggered as the reference for the subsequent recovery phase, rather than blindly using the equipment's rated nameplate parameters, avoids excessive injection of redundant current into the system after power restoration that exceeds the user's current actual needs. This makes the power recovery process settings more closely match the actual energy consumption curve of the actual business scenario.

[0015] In a second aspect, an artificial intelligence-based intelligent power regulation system for distribution boxes is also provided, including a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to implement the artificial intelligence-based intelligent power regulation method for distribution boxes as described in any of the above embodiments.

[0016] Beneficial effects: By integrating the multidimensional electrical characteristics of branch circuits with the local micro-area ambient temperature of circuit breakers, and relying on lightweight artificial intelligence classification models and communication probe technology, flexible and adjustable loads can be dynamically and accurately identified. At the same time, a multidimensional thermal contribution quantitative evaluation index based on the spatial adjacency relationship of circuit breakers is constructed to accurately identify the primary responsible branch that causes local thermal coupling and implement smooth derating. This achieves progress from macroscopic blind power outages to micro-area fine-grained derating, effectively preventing the aging of electrical equipment insulation and the hidden dangers of thermal damage, while maximizing the continuous power experience for end users. Attached Figure Description

[0017] Figure 1This is a schematic flowchart illustrating a power regulation method according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the branch attribute identification and thermal contribution analysis based on an artificial intelligence classification model according to an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the execution process of a dual-threshold temperature control strategy for adjacent circuit breaker groups according to an embodiment of the present invention. Detailed Implementation

[0018] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0019] like Figure 1 As shown, S101: Data acquisition and preprocessing.

[0020] In this embodiment, the main control module simultaneously activates two independent acquisition channels in each sampling cycle. First, the Hall current sensor and voltage sampling circuit in each branch intelligent circuit breaker node continuously output transient current and transient voltage at a first preset sampling frequency. Then, the main control module identifies the voltage fluctuation cycle of the power grid in real time by performing zero-crossing algorithm detection or phase locking on the transient voltage, thereby determining the current power frequency cycle. Second, the negative temperature coefficient thermistor attached to the gap between adjacent circuit breakers outputs the local ambient temperature at a second preset sampling frequency.

[0021] Subsequently, the control module performs filtering on transient current and transient voltage to suppress interference, and extracts the RMS current characteristics and multi-path harmonic component characteristics within each power frequency cycle based on the previously determined power frequency cycle, thereby constructing a harmonic feature vector corresponding to each power frequency cycle. Next, a continuous power frequency cycle of a preset length is used as a time window, and it slides with a preset step size. The statistical time sequence of the RMS current within the window and each harmonic component are spliced ​​along the channel dimension to construct a sliding time window tensor reflecting the time-frequency characteristics. The preset length is 100 power frequency cycles, and the preset step size is 1 power frequency cycle. Then, the sliding time window tensor, along with the timestamp, branch number, and local ambient temperature, is encapsulated into a standardized data frame and stored in the buffer for subsequent use.

[0022] As mentioned above, when any Hall current sensor or thermistor experiences a data out-of-bounds event, such as the current exceeding the upper limit of the range or the temperature reading being a negative saturation value, or there being no data update for three consecutive sampling cycles, the main control module immediately marks the branch as abnormal, freezes the data frame of the branch, and reports the event until the next sampling does not result in a data out-of-bounds event and there are no data updates for three consecutive sampling cycles, at which point the freeze is automatically lifted.

[0023] It should be noted that a branch circuit is an independent power supply circuit defined by the internal hardware wiring structure of the distribution box. Specifically, after the mains power is introduced through the main input terminal of the distribution box, it is split into several independent electrical output paths through internal busbars or wires. Each path that independently supplies power to a specific end-user device is thus established as a branch circuit.

[0024] S102: The standardized data frame is sent into the artificial intelligence classification model to complete the identification of the load equipment type, and the rigid and flexible attributes of each branch are dynamically labeled according to the handshake results.

[0025] After processing in step S101, the obtained standardized data frames are sequentially fed into a one-dimensional convolutional classification network pre-installed in the main control module according to their branch numbers. This network consists of three one-dimensional convolutional layers, two max-pooling layers, one flattening layer, and two fully connected layers. The kernel lengths of the three convolutional layers are 3, 5, and 7, respectively, and the number of channels are 16, 32, and 64, respectively. The output dimensions of the fully connected layers are 128 and 8, respectively. The last layer is followed by Softmax activation to output the probabilities of each category. In this embodiment, the output categories correspond to lighting equipment, data servers, fixed-frequency air conditioners, variable-frequency air conditioners, electric vehicle charging piles, electric water heaters, induction motor loads, and other categories.

[0026] The classification network was trained offline before the distribution box left the factory. For the eight load types mentioned above, standardized data frames were generated under five operating conditions: rated load, startup, half-load, full load, and frequent start-stop. Each type accumulated at least 10 hours of data, and the training, validation, and test sets were divided in a 7:2:1 ratio. The Adam optimizer was used iteratively with multi-class cross-entropy loss as the objective, achieving a classification accuracy of at least 95% on the test set. The convolutional network used here is only one implementation path; any classification method that can extract equipment categories from current timing and harmonic characteristics, such as residual networks, can be applied.

[0027] Then, based on the identification results, the local load attribute library is queried to obtain the default rigid / flexible attributes: devices with internal power electronic converters that can respond to external power regulation protocols, such as charging piles, variable frequency air conditioners, electric water heaters, and energy storage inverters, are marked as flexible loads by default; devices without adjustable interfaces, such as lighting, fixed frequency refrigerators, and data servers, are marked as rigid loads by default. Further, the main control module initiates a protocol handshake message to the identified branch end device. If a valid response is received within a preset timeout period, it is confirmed as a flexible load and its available protocol type is recorded. A valid response means receiving a feedback message within the timeout period that conforms to the preset protocol frame format, has a correct CRC check, and whose device feature code matches the identification category; otherwise, even if the attribute library marks it as flexible, it is temporarily downgraded to rigid processing. The preset timeout period is 500ms. Thus, each branch obtains an attribute label that dynamically updates with the communication status, enabling subsequent adjustment decisions to accurately distinguish between adjustable and non-adjustable loads.

[0028] S103: Based on the spatial adjacency relationship of circuit breakers in the distribution box and dual temperature thresholds, identify the thermal risk level of the current adjacent group.

[0029] Specifically, the topology configuration file of the distribution box is read to obtain the physical position sequence number of each smart circuit breaker on the guide rail, and a spatial adjacency table of the circuit breakers is constructed accordingly. This table uses a sliding window length of 3 as the default, that is, any three consecutive parallel circuit breakers are regarded as an adjacent group. It can also be expanded or reduced in the range of 2 to 5 to adapt to different density guide rail arrangements.

[0030] Then, two independent temperature thresholds were set for each adjacent group, corresponding to two different physical failure mechanisms. The first temperature threshold is the false trip protection threshold, with a value of 60℃, which can be adjusted within the range of 55 to 65℃. It is based on the ambient temperature sensitivity curve of the bimetallic thermomagnetic trip unit. The second temperature threshold is the insulation aging protection threshold, with a value of 108℃, which can be adjusted within the range of 95 to 110℃. It is obtained by subtracting a 10℃ safety margin from the critical surface temperature of the circuit breaker sample after being continuously loaded at 1.13 times the rated current and 40℃ for 168 hours.

[0031] Finally, the real-time temperature output by the thermistors in the adjacent group is compared with two temperature thresholds: when the real-time temperature is below the first temperature threshold, the current power supply is maintained; when the real-time temperature reaches or exceeds the first temperature threshold but does not reach the second temperature threshold, it is determined that there is a risk of false tripping of adjacent circuit breakers, and the system enters the normal suppression process, which involves reducing the rated power by 10% or reducing the current by 2A; when the real-time temperature reaches or exceeds the second temperature threshold, it is determined that the insulation aging threshold has been approached, and the system enters the aggressive suppression process, which doubles the suppression step size of the normal process, such as reducing the rated power by 20% or reducing the current by 4A. Through the above dual-threshold determination, the system can respond gently in the case of mild overheating and quickly suppress the heat source in the case of severe overheating, avoiding the strategy mismatch caused by covering two physical mechanisms with a single threshold.

[0032] S104: Calculate the thermal contribution of each branch in the adjacent trigger group and generate a derating command with protection constraints according to the principle of flexibility priority.

[0033] For the adjacent groups determined to be triggered in step S103, the thermal contribution of each branch within that adjacent group is calculated to quantify its actual contribution to the temperature rise of the adjacent micro-area. The motivation for constructing this index is that instantaneous current amplitude alone is insufficient to distinguish between the three distinct heating principles: steady-state Joule heating, fluctuating shock heating, and start-stop pulse heating. Therefore, this step uses the following weighting formula: In the formula, branch road thermal contribution branch road The effective value of the current, branch road The standard deviation of the effective value of the current within the sliding window, The number of times branch i starts and stops within one minute; , , The weighting coefficients are dimensionlessly normalized and have values ​​of 0.5 W / A, 0.8 W / A, and 0.5 W / (times·min). - ¹).

[0034] Subsequently, the adjacent branches are sorted in descending order of thermal contribution, and the branch with the highest thermal contribution and the attribute of flexible load is selected as the primary suppression target; if the first branch is a rigid load, the next highest flexible load is selected; if there are no flexible loads in the entire group, an alarm is output and the system reverts to power-off as a fallback. A target limiting current is generated for the selected branch: In the formula, The target current limit for the selected branch, i.e., the derating command. The effective value of the current in the selected branch; The derating step size is set to 10% of the rated current. If the adjacent group to which this branch belongs is in an aggressive suppression process, the value is doubled, i.e., 20% of the rated current. This is the minimum operating current limit for the selected branch.

[0035] S105: The adjustment command is sent to the end flexible load through protocol adaptation and closed-loop feedback is completed.

[0036] It is important to note that before issuing the adjustment command, the effective value of the instantaneous current before the suppression action is triggered needs to be stored in a cache as a reference current. Then, the main control module sends the derating command generated in step S104 to the corresponding smart circuit breaker node via the communication bus. The protocol conversion module within the node translates the unified format command into the native protocol of the end device. For example, the charging pile side uses OCPP, the air conditioner side uses Modbus, and some electric water heaters that only support PWM directly output the modulation signal. The controller of the end flexible load adjusts the duty cycle of its internal power electronic converter accordingly, effectively reducing the current drawn from the distribution box. At the same time, the mechanical tripping mechanism remains inactive, and the circuit is physically connected. Subsequently, the smart circuit breaker node sends back the measured effective value of the current. The main control module compares the measured value with the target value, and determines that the suppression is in place when the deviation is less than the convergence threshold.

[0037] When the real-time temperature drops below the first temperature threshold (a preset multiple) and remains below it for 30 seconds, the system gradually increases the current with a preset recovery step size until it returns to the reference current before suppression. The preset multiple is 0.85 times, and the recovery step size is preset to 5% of the rated current. If the temperature rises again during the recovery process and touches the first temperature threshold, the recovery will stop immediately and the current will return to the previous level.

[0038] If the flexible load fails to reduce its capacity as instructed, causing the temperature to continue rising to the second temperature threshold, the system immediately activates the mechanical trip to forcibly disconnect the branch, thus awaiting subsequent personnel intervention. After completing this round of adjustment, the system returns to step S101 for a loop.

[0039] The present invention also provides an intelligent power regulation system for a distribution box based on artificial intelligence. The system includes a processor and a memory, the memory storing computer program instructions. When the processor executes the computer program instructions, it implements the intelligent power regulation method for a distribution box based on artificial intelligence according to the first aspect of the present invention.

[0040] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and therefore will not be described in detail here.

[0041] like Figure 2As shown, the horizontal axis represents the specific load type identified, such as charging piles and air conditioners; the vertical axis represents the quantified value of each branch's contribution to temperature rise, i.e., the thermal contribution. The figure clearly distinguishes between flexible and rigid loads through color and text, which proves that the system can accurately identify which branches can participate in regulation and which branches must prioritize supplying rigid loads.

[0042] like Figure 3 As shown, a red temperature curve rises over time. The figure shows two distinct temperature thresholds. When the temperature enters the two different shaded areas, the system triggers conventional and aggressive suppression actions, respectively. This demonstrates that the system does not simply shut off the power, but rather takes gentle interventions of varying intensities based on physical mechanisms.

Claims

1. An artificial intelligence based intelligent power regulation method for distribution box, characterized in that, include: S101: Collects the current and voltage of each branch and the local temperature between the casings of adjacent smart circuit breakers. Based on the current and voltage, it obtains the effective value of the current and multiple harmonic components, and together with the local temperature, it forms a standardized data frame. S102: Input the standardized data frame into the classification model to obtain the load type; dynamically label each branch as a flexible load or a rigid load according to the load type and communication handshake results; S103: Determine at least two consecutively installed smart circuit breakers on the guide rail as adjacent groups, and set a first temperature threshold for each adjacent group; In response to the local temperature exceeding the first temperature threshold, calculate the thermal contribution of each branch in the group. The thermal contribution is positively correlated with the square of the effective current value, the standard deviation of the effective current value within the set window, and the number of start-stop cycles per unit time. S104: Select the branch with the highest heat contribution and marked as a flexible load as the regulation target, and send a derating command to it; the derating command includes the target limiting current. S105: The terminal device reduces its own current accordingly until it returns to the reference current before suppression.

2. The intelligent power regulation method for distribution boxes based on artificial intelligence according to claim 1, characterized in that, Also includes: A second temperature threshold is set for each adjacent group, and the second temperature threshold is higher than the first temperature threshold. In response to the local temperature exceeding the second temperature threshold, an aggressive suppression process is initiated, in which the target current reduction of the derating command is twice that of the conventional suppression process.

3. The intelligent power regulation method for distribution boxes based on artificial intelligence according to claim 1, characterized in that, Restoring the reference current to its pre-suppression state includes: When the local temperature is lower than the product of a first temperature threshold (a preset multiple) and 0.85 and this condition persists for the duration of the cooling period, the recovery process is initiated, gradually increasing the current with a preset recovery step size until the reference current is reached.

4. The intelligent power regulation method for distribution boxes based on artificial intelligence according to claim 3, characterized in that, Also includes: If the local temperature exceeds the first temperature threshold again during the recovery process, the recovery will be immediately paused and the current value will be returned to the previous level.

5. The intelligent power regulation method for distribution boxes based on artificial intelligence according to claim 1, characterized in that, The communication handshake process includes: The main control module sends a protocol handshake message to the identified branch end device. If it receives a response message containing the correct device identifier and passing CRC check within a 500ms timeout period, it determines that the device is a flexible load and records its available protocols; otherwise, it temporarily downgrades the device to a rigid load.

6. The intelligent power regulation method for distribution boxes based on artificial intelligence according to claim 2, characterized in that, It also includes a safety net mechanism: An alarm is triggered when the local temperature of any adjacent group exceeds the first temperature threshold but there is no flexible load in the group. When any flexible load fails to adjust according to the derating command, causing the local temperature to continue to rise to the second temperature threshold, the mechanical tripping mechanism of the corresponding smart circuit breaker is activated to forcibly disconnect the branch.

7. The intelligent power regulation method for distribution boxes based on artificial intelligence according to claim 1, characterized in that, The classification model consists of three one-dimensional convolutional layers, two levels of max pooling layers, one flattening layer, and two fully connected layers.

8. The intelligent power regulation method for distribution boxes based on artificial intelligence according to claim 1, characterized in that, The target limiting current needs to satisfy the following relationship: ； In the formula, The target current limit for the selected branch is the derating command; The effective value of the current in the selected branch; The minimum operating current limit for the selected branch; To reduce the volume step size.

9. The intelligent power regulation method for distribution boxes based on artificial intelligence according to claim 3, characterized in that, Methods for obtaining the reference current include: Before issuing the adjustment command, the effective value of the instantaneous current before the suppression action is triggered needs to be stored in the cache as the reference current.

10. An intelligent power regulation system for a distribution box based on artificial intelligence, comprising a processor and a memory, characterized in that, The memory stores a computer program, and the processor executes the computer program to implement the intelligent power regulation method for distribution boxes based on artificial intelligence as described in any one of claims 1-9.

Images